Housing Profile and Recommendation for Housing Provident Fund Using Two-Tower Neural Collaborative Filtering

The housing provident fund system has provided tremendous assistance in helping provident fund users take out loans for home purchase services since its inception. Currently, as digitalization and information technology continue to strengthen, the country emphasizes the need to use digital technology (e.g., big data, artificial intelligence, etc.) extensively in government management services and to drive innovation in housing fund management, service, and supervision. Profile and recommendation technologies were mostly used in the field of financial technology, such as e-commerce platforms and commercial banks, but we innovatively apply the big data profile and artificial intelligence recommendation methods in the provident fund. We propose a multi-dimensional profile model based on the big data from the Housing Provident Fund Center and establish a housing profile labeling system to help the center in digital management better. At the same time, based on the two-tower neural network collaborative filtering algorithm, we built a housing recommendation model based on the past purchase transaction history, which can be used to improve the digital service level of the center. After experiments, our model is much more accurate than the compared traditional collaborative filtering algorithms. Through our study, the Housing Provident Center can enhance digital management techniques, improve service level, and promote intelligent government construction.